A Data Framework to Understand the Lived Context for Dementia Caregiver Empowerment Marta Belay, Charles Henry, Daran Wynn, Tonya Smith-Jackson Department of Industrial and Systems Engineering, NC A &T State University, Greensboro, NC 27411 mwbelay@aggies.ncat.edu, tlsmithj@ncat.edu Abstract (Alzheimer’s Association, 2014). According to the World Agitation in dementia patients is characterized by several Health Organization (WHO) statistics, about 35.6 million features, such as physical and verbally aggressive and non- people are affected by dementia worldwide, and Alz- aggressive behaviors. Such behaviors affect not only the heimer’s disease contributes to 60-70% of the cases patients, but also their caregivers’ quality of life. The onset (WHO, 2012). of agitated behaviors can be unpredictable and can also be influenced by environmental factors, which introduce Agitation is a common and challenging consequence of challenges to caregivers when caring people with dementia dementia, which occurs in 90% of the patients (Colombo et (PWD). The purpose of this study is to analyze multiple al., 2007). Various stages of dementia require different sets forms of qualitative and quantitative data obtained through of skills from caregivers, and most caregivers do not have behavioral and environmental sensors. Data about body training in possible interventions. This results in stress and gestures, activity and task sequences, ambient light, sound and temperature will be obtained. Caregiver logs and increased caregiver burden and also leads to institutionali- medical history from nurses and psychiatrists are the sources zation of patients in long term care facilities (Steinberg et of qualitative data. Data framework will be used to collect, al., 2008). In addition, it incurs higher economic cost to structure, extract, analyze, interpret and integrate various provide the necessary care for a person with dementia formats and large amount of data. This approach helps to (PWD). The cost of dementia care in 2010 was estimated conceptualize the lived context of PWD. The information discovered will be used to generate trained models to to be between $157 billion and $215 billion by a nation- identify the patterns of agitation associated with the wide study (Hurd et al., 2013). In 2013, the estimated eco- environmental factors. It will also be used to develop a nomic value of care provided by unpaid caregivers was monitoring and dashboard system so caregivers and $220.2 billion. Similarly, aggregate cost of care provided healthcare providers can understand and avoid with payment was $214 billion (Alzheimer’s Association, environmental triggers. The research outcome will provide cost effective technology to reduce or prevent agitation in 2014). dementia. Empowering caregivers to reduce stress and agitation in PWD will have positive impacts on the PWD, the caregiv- er, and the associated cost of care can be reduced. The fol- 1. Introduction lowing is a case scenario describing the experience of a Dementia is a general term, which describes conditions caregiver attending her mother from Alzheimer’s associa- characterized by decline in memory or cognitive function tion webpage (www.alz.org). that affects a person’s ability to perform day-to-day activi- ties (Alzheimer’s Association, 2014). The occurrence rate “I’ve been the primary caregiver for my mother with de- of all types of dementia among individuals older than 71 mentia/Alzheimer for the past nine years. She’s 86 and is was 13.9% in 2002 (Plassman et al., 2007). This rate corre- fading away by inches and by bits and pieces. It is so unbe- sponds to 3.4 million individuals in the USA. The preva- lievably cruel and torturous to watch someone who was an lence rate of dementia has been found to increase with age excellent teacher and active lover of life be whittled away from 5% of those aged between 71 and 79 years to 37.4% by this hideous disease a tiny bit at a time. I’m convinced of those aged 90 and older (Plassman et al., 2007). The she contracted it through hormone replacement therapy, most common cause of dementia is Alzheimer’s disease. In which she had for too long and past the age of 75. I really the United States, an estimated 5.2 million people have don’t know how to convey how horrible this is for her and Alzheimer’s disease and it is estimated that in every 67 for me. She has suffered more than we can ever know, both seconds someone develops the disease. By the mid- physically and mentally. I have given 20 percent of my life century, the occurrence is estimated to be every 33 seconds to caring for her 24/7. Predictably, my life has received no attention at all. I have no husband, no family, no career, of the data; where data forms, types, sources, and scales no retirement, and no plans for the future. I’ve had to en- vary extensively. This paper describes the frameworks that dure my own personal heartaches in silence, including los- serve as taxonomies and ontologies to assist our research ing several beloved pets over the years, losing relatives team to plan, collect, extract, analyze, and interpret the and my own battle with skin cancer. Everything is second- multiple data streams from the BESI project. As the re- ary when you are a caregiver. Your life is forfeited, and search is at its early stage, the conceptual data framework because this battle cannot be won, you will ultimately fail. which facilitates the data collection and analysis, and There is simply no way to put a good face on this experi- which also forms the basis for the advancement of the ence.” technology is presented in this paper. Such stories are common among caregivers. Caregivers of PWD have a 50% chance of experiencing depression due to the stressors they experience with the changed be- 2. Literature Review havior, unpredictability, reduced cognitive abilities and In modern data-intensive science, more consideration has role changes (Schulz et al., 1995). Caregivers with depres- been given to the challenges of handling massive data for- sion have increased morbidity and mortality (Pruchno and mats and volumes. Considering the data ecosystem as a Potashnik, 1989), and PWD in these dyads have shorter whole is very essential to truly address the challenges of times before institutionalization (Schulz et, al., 1999). In- very diverse multidisciplinary data. Understanding com- stitutionalization may be linked to a more rapid psycholog- plex system problems involving heterogeneous and diverse ical decline, since the individual is placed in an unfamiliar interdisciplinary research data requires mixed data integra- environment at a critical period and becomes cared for by tion and analysis (Parsons et al., 2011). A conceptual data individuals they do not know. framework can be used to map the relationships and de- To address this significant challenge, a research team, pendencies among various scientific data sources, types of comprised of investigators at NC A&T State University, data produced and used, and curation activities associated University of Virginia and the Carilion Center for Healthy with the data (Cragin et al., 2010). Conceptual mapping of Aging, has focused on the goal to identify engineering- data frameworks can also be used to reduce qualitative da- based interventions to increase caregiver empowerment ta, analyze themes and interconnections in the data through the use of tools to predict and minimize agitation (Onwuegbuzie et al., 2009). Data frameworks are helpful episodes among PWD. The envisioned system, Behavioral to identify types of data to be collected and data analysis and Environmental Sensing and Intervention (BESI), is a techniques to be used (Parsons et al., 2011). They also complex cyber-socio-physical system that incorporates serve as aids to develop new methods of analysis. technologies, social dyads and contexts. In other words, the The three V’s (volume, variety and velocity) are charac- Cyber-socio-physical system is consisted of three subsys- teristics of big data. Volume refers to the large amount of tems which comprise of various components. This complex data, variety refers to different types of data and velocity system will be used to acquire multiple forms of descrip- stands for the rate of data accumulation (Berman, 2013). tive data to build a knowledge base of the ecosystem sur- The greatest benefit of big data is the ability to link seem- rounding agitation. Data will be analyzed to understand the ingly different disciplines for the purpose of developing lived context of a PWD and to develop a model that can be and testing hypotheses that cannot be approached within a used to predict agitation events associated with the envi- single knowledge domain. With few exceptions, big data is ronmental conditions. A monitoring system which recog- ordinarily analyzed in incremental steps; the data are ex- nizes agitation epochs will be developed to send real time tracted, reviewed, reduced, normalized, transformed, visu- notification for caregivers. Secured web-based interface alized, interpreted and re-analyzed with different methods monitoring system will be used to display the sensor data (Bari et al., 2014). Big data has many implications for pa- for health care providers, caregivers and other authorized tients, healthcare providers, researchers and other users. The web-interface will be refined with input from healthcare constituents. It will also impact how these play- nurses, caregivers, and health informatics to ensure it is us- ers engage with the healthcare ecosystem, especially when er friendly and easily interpreted. Sensor data will be external data, regionalization, mobility and social network- grouped by category such as physical agitation, tempera- ing are involved (Murdoch and Detsky, 2013). ture and noise level, and other environmental stimuli. Us- Data mining, knowledge extraction, information ers can further navigate the interface to view data from in- discovery, information harvesting and data pattern dividual sensors. processing are some of the names used in the past to refer As a result, caregivers can intervene on the PWD and the to the process of finding useful patterns in data (Fayyad et environment before agitation escalates. BESI will be an al., 1996), also known as knowledge discovery. Fayyad et empowering tool for caregivers of PWD with cost effective al. (1996) define knowledge discovery as a series of solution. Yet, the challenge of BESI lies in the immensity activities for making sense of data. They distinguish data data structuring and analysis on complex systems. The in- mining as a specific step in the knowledge discovery vestigators are developing a cyber-socio-physical system to process which focuses on the application of certain assist caregivers and providers in the management of agita- algorithms to extract useful information (knowledge). In tion in dementia. The cyber-socio-physical system is a contrast to these distinct views of knowledge discovery and complex system based on its characteristics – interrelated- data mining, Peng et al. (2008) use combined process of ness, autonomous components, and dynamic. data mining knowledge discovery (DMKD). They define The study to be conducted will use a remote ethnograph- DMKD as extraction of useful information (knowledge) ic approach to collect data about the physical agitation of a from data and this extraction is achieved by learning new PWD and the natural living environments of the PWD and methods and techniques. These methods and techniques are caregivers. This is achieved by making use of different used in the pre-processing and post processing of data, sensors on the patient as well as the surrounding environ- specifically for discovering previously unknown patterns ment. Body-worn sensors are placed on the PWD, which and building predictive models from the data (Peng et al., capture the movement of the patient at multiple parts of the 2008; Maimon et al., 2010). body to detect different stages of physical agitation. Envi- Previous works which focus on monitoring agitation be- ronmental acoustic sensors are installed to capture infor- haviors were reviewed. Bankole et al. (2012) conducted a mation about ambient noise and speech features. Light and study to explore the ability of a custom inertial wireless temperature sensors are used to measure ambient environ- body sensor network (BSN) to detect and quantify agita- mental conditions. Additional set of motion sensors are in- tion. The initial study was focused on validating the BSN. stalled near doorways to detect movement from one room The research work consisted of data collection on selected to another. The sensor networks, wireless devices and lap- subjects at different times of the day. From assessment of top-based stations (physical structures), algorithms and the pilot results, it was concluded that the BSN was a valid computations constitute the cyber subsystem. measure of agitation. The ability of the BSN for continuous Different subjective measures are used to recognize the and real-time monitoring was also examined (Bankole et agitation events, frequency, type, and stress level experi- al., 2011). enced by the caregivers. These measures will help to quan- In summary, a data framework is used to plan, structure, tify agitated behaviors and their impact on the PWD and and organize different data formats and large amounts of caregiver separately as well as on the dyad as a unit. PWD, data for data analysis and data integration in caregiver-healthcare providers, the PWD-Caregiver dyad, multidisciplinary research. It helps to integrate, process, patient’s family and friends make up the social subsystem. visualize, and present data in a meaningful way. Agitation is influenced by a number of environmental fac- Knowledge discovery processes are implemented to tors such as ambient temperature, sound and light level, prepare, select and cleanse data. Proper interpretations of social density etc. It is important to track knowledge of this mined data from the research domain are possible using environment which makes up the physical subsystem to these processes. Constructing an integrated and interactive minimize the occurrence of agitation events in the patients. data framework with the application of knowledge The problem space addressed by this research is three discovery and data mining will provide a map of mixed fold: analytical landscape for multidisciplinary researchers. This 1. The volume and variety of the data requires an organ- data framework can also facilitate research team izing data framework that guides input, structuring, communication, collaboration and the development of and analysis of the various forms of data shared mental models. Most importantly, data frameworks 2. The complexity of the system (cyber-socio-physical) support reasoned action when analyzing data. If requires a data framework to organize team members’ frameworks are organized and agreed upon ahead of time integrated mental models as the system is developed while researchers are focusing on the primary research from concept to final prototype questions and objectives, the analytical processes and 3. The data framework is needed to facilitate the data to reasoning from the data will be more aligned with the line design translation process to achieve the final out- of inquiry established by the problem to be addressed and comes to benefit caregivers and PWD the research goals. In this way, research integrity is maintained. 4. Data Framework Development Process and Results 3. Purpose of the Research Developing a conceptual framework for a specific study The purpose of this research is to develop and implement a incorporates a system of concepts, assumptions, expecta- data framework for the research and design team to apply tion, beliefs and theories that support the research Identify the Identify the data Data extraction from subsystems of Obtain correlation source and acquisi- each subsystem and the cyber- between extracted tion mechanism for Identify the data for- socio-physical data each subsystem mat system Figure 1. Flow chart that shows processes followed to generate data framework (Wang et al., 1995). Idea association can be regarded as the in the BESI system. There are individual social sub- catalyst that facilitates the interaction among researchers systems, dyadic social sub-systems (i.e., PWD-caregiver; and design participants. By linking the researchers’ and de- caregiver-healthcare provider), and group-level subsystems signers long term memory internally and previous partici- (more than two individuals). pant knowledge externally, diverse design ideas can be The Physical/environmental subsystem consists of envi- generated (Lai and Chang, 2006).In the BESI project, a ronmental conditions that surround the dementia patients. team that consists of multidisciplinary experts from com- Temperature, sound and light intensity, physical movement puter and electrical engineering, human factors and ethnog- and speech features represent the physical subsystem. Am- raphy, geriatric psychiatry, and nursing conducted a brain- bient conditions and gross movement data are gathered us- storming session to enhance their previous knowledge ing environmental and door way sensors. Various formats about the cyber-socio-physical system with additional in- (i.e. relative frequencies of codes, ratio, categorical, bina- novative perspectives. Individual ideas were linked with ry) of the desired information are extracted from the col- greater technical depth to generate the following flow lected data using algorithms. chart. The flow chart shows the basic steps followed to The researchers can use the following data framework to generate the integrated and inclusive data framework. plan and organize their data collection, extraction, and Figure 1 demonstrates the data framework of the BESI analysis activities. For instance, to monitor the behavior of project. This data framework accounts for the interactions a PWD and their environment, body-worn and environ- of the components of the cyber-socio-physical subsystems. mental sensors are used. These sensors continuously pro- Cyber-socio-physical systems are comprised of three sub- vide data about the physical movement of the patient and systems. The cyber subsystem consists of the inertial body- ambient conditions in the room. However, sensory raw da- worn sensors, environmental sensors, wireless Bluetooth ta is often difficult to understand and interpret, especially devices and computers. The sensor stream provides contin- when sensory data comes from multiple sensors. It should uous data about body motions and ambient living space be noted that the sensor data is collected every second and conditions. Similarly, the wireless Bluetooth gives infor- when this frequency of data collection is repeated for mul- mation about the location of person in a house in different tiple sensors, the researchers would be facing challenge of times of the day and night. A base workstation communi- handling and interpreting large amounts of data. Therefore, cates with all the sensors and wireless devices. Data extrac- researchers should divide the complex system into subsys- tions from the sensory devices are done by applying differ- tems and then apply the data framework to extract and in- ent signal processing algorithms. The extracted data will terpret the data from each subsystem independently. have different format such as binary, continuous, ratio. In addition, the data interpreted from each subsystem The social subsystem encompasses the dementia patient, should be integrated and correlated with each other to have caregivers, nurses, patient’s family and friends, health care knowledge of the whole system behavior. The sensory data providers. Interviews, caregiver diaries, assessment batter- from the cyber and physical subsystems, for instance, ies are used to collect data. The collected data provides in- should be converted to meaningful form to identify pat- formation about behavioral pathology in dementia patients, terns of movement which allow categorization of the pa- cognitive level, aggressive and non-aggressive agitation tient’s behavior and the environmental conditions respec- symptoms, dementia stage, functional capacity, sleep tively. This can be achieved by simultaneous collection quality, quality of life for the caregivers and patients, etc. and separate interpretation of the data from both subsys- Content analysis and score calculations are applied to filter tems. The interpreted data are then integrated to identify useful data from the collected information. It is important the environmental condition which contributes to the agita- to understand the various levels of social subsystems with- tion. Figure 2. Cyber-socio-physical system data framework 5. Discussion and Conclusion ta sets to be collected and analyzed (Table 1). This table A complete data framework provides researchers the ad- was generated through mock-up data set that is currently vantage of dealing with complex systems in several ways. under development. The time scales, formats, sources, and It enables identification of subsystems of a complex system numerical scales will differ. The large volume of data, fre- and helps to identify individual subsystem data sources and quency of data collection, and variety of the data which data acquisition mechanisms. It provides a platform to- come from multiple sensors demand data framework that wards extraction of useful information from the different guides inputs, structuring, extraction, and analysis. This subsystems. It helps to correlate the data from the subsys- framework will also evolve more as the research team pro- tems which are useful for understanding of the entire sys- ceeds through the different project phases. tem which in our research is definition of the lived context Verification and validation of basic BESI sensing and of PWD’s. The project will have very diverse and large da environmental assessment will be done in a controlled set- up and in the homes of PWD’s. Reliability of body-worn Table 1. Example of sensor data with time stamps: BSN and TM represent body worn and environmental sensor ID. Sensor ID Time stamps Desired information Data type BSN 001 11/06/2015 04:45:34 Pre-agitation, agitation event Interval, continuous BSN 005 11/06/2015 04:50:56 Agitation period, post agitation Ratio level, continuous TM 008 11/06/2015 17:15:25 Temperature level Ratio level, continuous MS 007 11/06/2015 23:12:55 Social density Binary Caregiver diary 11/06/2015 07:05/05 Agitation frequency, level, Relative frequency of themes, caregiver self-reports of codes, quantitative ratings psychosocial variables (qualitative themes, continuous scale ratings) Assessment battery 12/07/2015 08:15:25 Level of cognition, functional Ratio, interval, continuous capacity sensors is validated based on caregiver diaries and assess- ment batteries of agitation events. Data from caregivers Acknowledgement will be obtained through tablet diaries with structured This research is jointly funded by National Science prompts and a time-stamped report. This data is simultane- Foundation and National Institute of Health under one ously collected with body worn-sensor data. Constant Award number IIS-1418622. comparison, narrative and content analysis are used to ana- lyze qualitative data from caregivers, whereas one or more algorithms and statistical techniques are used to analyze References data from sensor streams. The results of quantitative analy- Alzheimer’s Association. Alzheimer’s disease Facts and Figures. sis of both quantitative and qualitative data are combined Includes a special report on women and Alzheimer diseases. at the interpretation level to validate the accuracy of sensor Alzheimer’s & Dementia Volume 10, Issue 2, 2014. activities. However, each data set remains analytically sep- Alzheimer’s association, Muffet’s Story arate from each other. Therefore, to plan the data collection http://www.alz.org/living_with_alzheimers_15512.asp and analysis processes at the intersection of the two sub- systems, it is important to use BESI data framework. Bankole, A., Anderson, M., Smith-Jackson, T., Knight, A., Oh, Data mining knowledge discovery (DMKD) approaches K., Brantley, J.S., Barth A.T., Lach, J. 2012. Validation of Non- will be carried out to extract useful information Invasive Body Sensor Network Technology in the Detection of Agitation in Dementia. American Journal of Alzheimer’s Disease (knowledge) from the each of the subsystems. Useful & Other Dementias 27(5):346-354. information from sensor streams will be analyzed to identify agitation epochs (pre-agitation, agitation, post- Bankole, A., Anderson, M., Knight, A., Oh, K., Smith-Jackson, agitation) and to model associated environmental T., Hanson, M. A., Barth A.T., Lach, J. 2011. Continuous, Non- conditions, which lead to slow or rapid perseverance of Invasive Assessment of Agitation in Dementia Using Inertial Body Sensors, Wireless Health, San Diego, USA. agitation episodes. Besides, sensor data will be combined with subjective data to support an integrated observation of Bari, A., Chaouchi, M., Jung, T. 2014 Predictive analytics for the social context of an agitation event. Finally, the dummies New Jersey John Wiley & Sons, Inc. analyzed data and the model developed will be used to inform design of monitoring system in which caregivers Berman, J.J. 2013 Principles of big data: Preparing, sharing, and analyzing complex information Waltham, MA: Elsevier Inc. and health care providers can get just in time notification about the agitation epochs and triggers. Colombo, M., Vitali, S., Cairati, M., Vaccaro, R., Andreoni, G., Data frameworks in multi-disciplinary research facilitate Guaita, A. 2007. Behavioral and psychotic symptoms of dementia team communication and exchange of information in the (BPSD) improvements in a special care unit: a factor analysis. process of understanding and analyzing complex systems. Archives of Gerontology and Geriatrics 44 (Supplement 1): 113- 120. It enhances knowledge and shared experience about which data analysis techniques can be integrated, how and why Cragin, M. H., Palmer, C. L., Chao, T. C. 2010. Relating data they are combined. In this regard, the BESI data practices, types, and curation functions: An empirically derived framework will provide an excellent perspective to the framework. Proceedings of the American Society for Information importance of multi-disciplinary research collaborations to Science and Technology. 47(1): 1–2. Pittsburgh, USA address critical societal needs. Daley, B. J, 2004. Using concept maps in qualitative research. Geriatric Psychiatry 23(2): 170–177. Concept Maps: Theory, Methodology, and Technology. Proceeding of the first international conference on Concept Trochim, W.M.K. 2006. Research methods knowledge base. Mapping. Pamplona, Spain. Retrieved on March 20, 2015. http://www.anatomyfacts.com/research/researchmethodsknowled Fayyad, U., Pitaetsky-shapiro, G., Smyth, P. 1996 Knowledge gebase.pdf Discovery and Data Mining: Towards a Unifying Framework. KDD-96 Proceedings. Wang, Y .R., Storey, C.V., and Firth, P.C. 1995. A Framework for Analysis of Data Quality Research. IEEE Transactions on Hurd, M. D., Martorell, P., Delavande, A., Mullen, KJ., Langa, K. knowledge and data engineering Vol. 7, no. 4. M. 2013. Monetary Costs of Dementia in the United States. The New England Journal of Medicine 368: 1326-1334. World Health Organization (WHO), Media Center: Dementia. Fact sheet No.362, Apr 2012. Lai, I. C. and Chang T. W. 2006. A distributed linking system for supporting idea association during the conceptual design stage. Design studies Vol 27, Issue 6, Pages 685–710. Maimon, O. and Rokach, L. 2010. Data Mining and Knowledge Discovery Handbook Second Edition. Springer. Murdoch, T. B. and Detsky, A. S. 2013. The Inevitable Application of Big Data to Health Care. The Journal of the American Medical Association Vol 309, No. 13. Onwuegbuzie, A. J., Dickinson, W. B., Leech, N. L., Zoran, A. G. 2009. A Qualitative Framework for Collecting and Analyzing Data in Focus Group Research. International journal of qualitative methods 8(3): 1-21. Parsons, M. A., Godøy, Ø., LeDrew, E., De Bruin, T. F., Danis, B., Tomlinson, S., Carlson, D.2011. A conceptual framework for managing very diverse data for complex interdisciplinary science. Journal of Information Science 37(6): 555–569. Plassman, B. L, Langa, K. M, Fisher, G. G., Heeringa, S. G, Weir, D. R., Ofstedal, M. B., Burke, J. R., Hurd, M. D., Potter, G. G., Rodgers, W. L., Steffens, D.C., Willis, R.J., Wallaceg, R.B. 2007. Prevalence of Dementia in the United States: The Aging Demographics, and Memory Study Neuroepidemiology, 29(1-2): 125-132. Peng, Y., Kou, G., Shi, Y., Chen, Z. 2008. A descriptive framework for the field of data mining and knowledge discovery. International Journal of Information Technology & Decision Making. Vol. 7, No. 4: 639–682. Pruchno, R.A. and Potashnik, S.L. 1989. Caregiving Spouses. Physical and mental health in perspective. Journal of the American Geriatric Society 37(8):697-705. Schulz, R., O'Brien, A. T., Bookwala, J., Fleissner, K. 1995. Psychiatric and physical morbidity effects of dementia caregiving: prevalence, correlates, and causes. Gerontologist 35(6):771-779. Schulz, R. and Beach, S.R. 1999. Caregiving as a risk factor for mortality: the Caregiver Health Effects Study. Journal of the American Medical Association 282(23):2215-2219. Steinberg, M., Shao, H., Zandi, P., Lyketsos, C. G., Welsh- Bohmer, K. A., Norton, M. C., Breitner, J. C. S., Steffens, D. C., Tschanz, J. T., and Cache County Investigators. 2008. Point and 5-year period prevalence of neuropsychiatric symptoms in dementia: the Cache County Study. International Journal of